Influence of childhood trauma and brain-derived neurotrophic factor Val66Met polymorphism on posttraumatic stress symptoms and cortical thickness

Interaction between childhood trauma and genetic factors influences the pathophysiology of posttraumatic stress disorder (PTSD). This study examined the interaction effect of childhood trauma and brain-derived neurotrophic factor (BDNF) Val66Met polymorphism on PTSD symptoms and brain cortical thickness. A total of 216 participants (133 healthy volunteers and 83 PTSD patients) were recruited. T1-weighted structural magnetic resonance imaging, BDNF rs6265 genotyping through blood sampling, and clinical assessments including the childhood trauma questionnaire (CTQ) and posttraumatic stress disorder Checklist (PCL) were performed. A moderated regression analysis, two-way multivariate analysis of covariance, and correlation analysis were conducted. An interaction between the CTQ and the BDNF polymorphism significantly influenced PTSD symptom severity. In fact, people with rs6265 Val/Val genotype and higher CTQ scores showed higher PCL scores. Additionally, this interaction was significant on both left fusiform and transverse temporal gyri thickness. Furthermore, the thickness of both brain regions was significantly correlated with psychological symptoms including depression, anxiety, rumination, and cognitive emotion regulation methods; yet this was mainly observed in people with the Val/Val genotype. The interaction between childhood trauma and BDNF polymorphism significantly influences both PTSD symptoms and cortical thickness and the Val/Val genotype may increase the risk in Korean population.


Results
Descriptive Statistics. The comparisons of demographic, genetic, and psychological characteristics between healthy and PTSD participants are provided in Table 1. The years of education, the childhood trauma questionnaire (CTQ), posttraumatic stress disorder checklist (PCL), posttraumatic growth inventory (PTGI), lifetime events checklist (LEC), hospital anxiety and depression scale (HADS), insomnia severity index (ISI), and ruminative response scale (RRS) scores, and some subscales of the cognitive emotion regulation questionnaire (CERQ) were significantly different between healthy participants and PTSD participants. However, since all variables, including both the CTQ and PCL scores, were normally distributed even for the total participants pool, further analyses with such a pool were executed, based on the assumption of normal distribution to represent the phenomenon in the general population.

Moderated Regression Analyses.
To examine the interaction effect of childhood trauma and rs6265 on the PCL score, moderated regression analyses were executed. The CTQ score was set as the independent variable since it represents genetic factors, whereas the rs6265 genotype (Val/Val vs. Val/Met + Met/Met) was set as a moderator. Firstly, the effect of demographic factors such as gender, age, and years of education, and the effect of the experience of traumatic events other than childhood trauma (LEC) were controlled for as covariates to reveal the pure effect of CTQ and rs6265 on PTSD symptoms. In the total participants pool, both the moderation model (R² = 0.355, p < 0.001) and the moderation effect were significant, given that the R² was increased due to the interaction (ΔR² = 0.020, ΔF = 6.492, p = 0.012). Although the rs6265 coefficient was not significant (B = 12.375, p = 0.060), coefficients of both the CTQ (B = 0.480, p < 0.001) and the interaction between the CTQ and rs6265 (B = −0.326, p = 0.012) were significant. Separately, this moderation effect was significant in PTSD participants (B = −0.353, p = 0.022), but not in healthy participants (B = −0.119, p = 0.466).
Secondly, another analysis was executed by including the effect of other psychological symptoms such as anxiety, depression, insomnia, and alcohol problems for additional covariates to exclude the effect of comorbid symptoms of PTSD. In the total participants pool, both the moderation model (R² = 0.757, p < 0.001) and the moderation effect were significant, given that the R² was increased due to the interaction (ΔR² = 0.005, ΔF = 3.954, p = 0.048). Although rs6265 coefficient was not significant (B = 7.690, p = 0.063), coefficients of both the CTQ (B = 0.236, p < 0.001) and the interaction between the CTQ and rs6265 (B = −0.161, p = 0.048) were significant. Separately, however, this moderation effect was marginally significant in PTSD participants (B = −0.212, p = 0.057), but not in healthy participants (B = −0.135, p = 0.302).
Results of both moderation analyses are described in Table 2. Although both analyses showed similar results, the second moderation model had much higher R² than that of the first moderation model (R² Model1 = 0.335; R² Model2 = 0.757) even with more covariates. Therefore, the second moderation model was accepted and displayed in Fig. 1.
As described in Table 2 and Fig. 1, the effect of the CTQ on the PCL was significant in those with the Val/Val (CC) genotype (B = 0.236, p < 0.001), whereas not in those with the Met allele (CT + TT) (B = 0.076, p = 0.157). As shown in Fig. 1, people with the Val/Val (CC) genotype showed lower CTQ score than those with the Met allele (CT + TT) when the CTQ score was below 47.500, while people with the Val/Val (CC) genotype showed higher CTQ score than those with the Met allele (CT + TT) when the CTQ score exceeded 47.500. Two-way MANCOVA. For further analyses, the total participants pool was used to represent the phenomenon in the general population, instead creating a division between the healthy and the patient groups. Among the total participants pool, the CTQ score was partitioned into two groups, at the point in which the interaction effect of the CTQ and rs6265 was observed, i.e. the score of 47.500. Subsequently, participants were divided into four following groups: (1) Group 1: 36 participants with low CTQ (below 47.500) and rs6562 Val/Val (CC) genotype, (2) Group 2: 99 participants with low CTQ and rs6265 Met allele (CT + TT), (3)Group 3: 31 participants with high CTQ (above 47.500) and rs6265 Val/Val (CC) genotype, and (4) Group 4: 50 participants with low CTQ and rs6265 Met allele (CT + TT).
The two-way MANCOVA was executed on cortical thickness in the total participants pool to examine the effect of low vs. high CTQ, rs6265 Val/Val vs. Met allele, and the interaction effect of the CTQ and rs6265. The demographic factors such as gender, age, and years of education were controlled for as covariates since they could affect cortical thickness, while the use of medication was not considered as a covariate since it was not associated with cortical thickness in previous studies 40,41 . The effect of the experience of traumatic events (LEC) was also controlled for as additional covariate to see the pure effect of the childhood trauma experience on cortical thickness. Considering that several cortical areas were analyzed as dependent variables simultaneously, the Bonferroni correction was used for multiple comparisons (p adj < 0.00125).
The    Correlation Analysis. A partial correlation analysis was executed on both left fusiform and transverse temporal gyri thickness to examine their relationship with psychological symptoms by generating 1000 bootstrapped samples for multiple comparisons 42 . Age was controlled for as a covariate since it was significantly correlated with both the thicknesses of the left fusiform (r = −0.217, p = 0.001) and transverse temporal (r = −0.470, p < 0.001) gyri, while gender (r = −0.036, p = 0.597; r = −0.006, p = 0.928, respectively) and education years (r = −0.057, p = 0.405; r = −0.002, p = 0.978, respectively) were not. As shown in Fig. 3 Table 2. The moderation effect of the rs6265 polymorphism on the relationship between the childhood trauma questionnaire (CTQ) and the posttraumatic stress disorder checklist (PCL) score while controlling for gender, age, years of education, lifetime events checklist (LEC), hospital anxiety and depression scale (HADS), insomnia severity index (ISI), and alcohol use disorders identification test (AUDIT). *p < 0.05, **p < 0.01, ***p < 0.001. www.nature.com/scientificreports www.nature.com/scientificreports/ transverse temporal gyri were significantly associated with the PCL score, both the anxiety and depression scores of the HADS, the RRS score, and the score describing catastrophizing in the CERQ. In addition, left transverse temporal gyrus thickness was significantly correlated with the score describing blaming others in the CERQ and the ISI score.
Further partial correlation analyses were executed on both the thicknesses of the left fusiform and transverse temporal gyri among people with the Val/Val (CC) genotype and the Met allele (CT + TT) for comparison. Again, age was controlled for as a covariate. As shown in Table 3, the correlations between cortical thickness and psychological symptoms differed among genotypes. In fact, the left fusiform gyrus thickness showed a significant negative correlation with PCL, HADS, RRS, and CERQ only in people with the Val/Val (CC) genotype. Similarly, the left transverse temporal gyrus thickness reported significant negative correlations with PCL, HADS, RRS, and catastrophizing in CERQ and a significant positive correlation with positive reappraisal in CERQ and PTGI only in people with the Val/Val (CC) genotype. In contrast, it indicated significant negative correlation with HADanxiety, blaming others and catastrophizing in CERQ in people with the Met allele (CT + TT).  www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
The current study aimed at examining the interaction effect of childhood trauma and BDNF rs6265 polymorphism on both the PTSD symptoms and cortical thickness among Korean population. People with both the Val/ Val genotype of BDNF rs6265 and a CTQ score over 47.500 showed severe PTSD symptoms. Furthermore, the interaction between this gene and childhood trauma also affected both the thickness of the left fusiform and transverse temporal gyri. Thickness, in turn, significantly correlated with symptoms severity, although the relationship differed between genotypes.
Our results revealed that the experience of childhood trauma could predict the severity of PTSD symptoms; however, this relationship was only significant in people with Val/Val (CC) genotype, but not in people with the Met allele (CT + TT) of the BDNF rs6265. In contrast, the BDNF rs6265 genotype did not affect PTSD symptom severity itself. Although previous studies revealed the effect of the rs6265 polymorphism not to be significantly related with PTSD, they did not consider the environmental effect 16,19,23,25,26,43,44 . Therefore, our results validated the hypothesis stating that gene-environment interaction may be the key to understanding the development of PTSD symptoms.
This moderation effect was significant both in the total participants pool (B = −0.161, p = 0.048), marginally in PTSD participants (B = −0.212, p = 0.057), but not in healthy participants (B = −0.135, p = 0.302). These group differences could be a result of dissimilarities in both the pathological characteristics and the genetic predisposition of our groups. For example, PTSD participants reported a higher CTQ mean score with a smaller standard deviation than healthy participants, which could imply the possibility of the floor effect. In addition, the frequency of Met carriers, the protective allele, was slightly higher in healthy participants (73.1%) than in PTSD participants (61.9%).
This interaction effect of childhood trauma and BDNF polymorphism was also found on the cortical thickness of both the left fusiform and left transverse temporal gyri. These regions were in fact thinner in participants with Val/Val genotype and with higher CTQ scores than in those with lower CTQ. Furthermore, both regions resulted thinner in participants with higher CTQ scores and with the Val/Val genotype than in those with the Met allele. Moreover, cortical thickness of both brain areas was negatively associated with scores of PTSD symptoms severity. These results suggest that the gene-environment interaction could influence the thickness of the left temporal regions, which are also related to PTSD symptoms.
Previous studies discovered a significant correlation between the left fusiform gyrus and PTSD [45][46][47] . The left fusiform gyrus is generally known to be associated with cognition, learning processes [48][49][50] , negative cognition of PTSD such as dysfunction of memory, dissociative symptoms 51 , and threat appraisal 47 . The current research is in line with these studies since the thickness of the left fusiform gyrus was significantly correlated with PTSD symptoms and other negative cognition (e.g., rumination, blaming others, and catastrophizing). Additionally, the left transverse temporal gyrus, also referred to as Heschl's gyrus, was described to be linked to PTSD in several studies [52][53][54][55][56] . The left transverse temporal gyrus appears to be involved in assigning a spectral order to pitch information 57 and language comprehension 58 , which could be related to the low-level perceptual deficits observed in www.nature.com/scientificreports www.nature.com/scientificreports/ PTSD 59 . Although this study did not evaluate low-level perception, our results highlight the importance of the left transverse temporal gyrus in PTSD.
Other than PTSD symptoms, this study found the relation of left fusiform and transverse gyri with diverse psychological symptoms including anxiety, depression, rumination, insomnia, and some subtypes of cognitive emotional regulation, which is in line with various other studies. The reduced fusiform gyrus thickness is known as a trait marker for vulnerability to depression 60 and social anxiety 61 . The fusiform gyrus is also considered to be related with the retrieval of memories and rumination 62 . In addition, the thin transverse temporal gyrus was related with depression 63 , and the decreased connectivity in the transverse temporal gyrus was found in high anxiety people 64 , in insomnia patients 65 , and in depressive patients with rumination 66 . Although transverse temporal gyrus is known to be related with exteroceptive sensory perception, a frontoparietal network including this area could be associated with cognitive control as well 67 . Therefore, results of this study support the possibility that these brain regions could act as key roles regarding psychopathology including PTSD and other various symptoms.
Results from this study suggest the BDNF rs6265 Met allele assumes a protective genotype when interacting with childhood trauma; in contrast, the Val/Val genotype may increase the risk of developing PTSD symptoms and cortical thickness. However, previous studies showed mixed results on this matter, as some indicated the Met allele to be risky 8,10,[20][21][22] , while others proposed the Val/Val genotype to be the risky one instead 27,28,33 . A possible reason behind such a difference may be participants' racial differences 26 . For example, our participants reported higher rs6265 Met allele (66.8%) than did other studies on western participants (approximately 31-40%) 68,69 . This said, the allele frequency found in our study participants was similar to the one reported by other studies (63.64-71.64%) conducted on Korean participants 23,70,71 .
A study with Korean participants revealed an absence of association between the rs6265 and PTSD 23 , while others found the Met allele to act as an allele increasing the risk of developing PTSD 71,72 . However, when considered the interaction of rs6265 and childhood maltreatment, the Val/Val genotype was reported to be associated with a higher anxiety than the Met/Met genotype in Korean females 39 . Furthermore, the participants with the Val/ Val genotype showed high harm avoidance traits after recent negative stressors in Korean participants 73 .
Another possible reason behind the mixed results related to the genotype effect may be gene-gene interaction. A few studies suggested in fact that the BDNF genotype may interact with the catechol O-methyltransferase (COMT) and the serotonin transporter gene linked promoter region (5-HTTLPR) to produce differences in resilience 74 and onset age of depressive disorder 75 , respectively. Further studies investigating such a gene-gene interaction are needed, given that PTSD is known to be related to resilience 76 , depression 77 , and other psychological factors.
Despite significant findings, there are some limitations in this study. Firstly, childhood trauma was assessed retrospectively. Although the CTQ is known to present a stable measurement over time 78 , additional longitudinal studies would be helpful to determine the effect of childhood trauma with accuracy. Furthermore, this study considered the BDNF polymorphism only. Multi gene effects should be explored for better define the gene-environment interaction in future studies, including genome wide association studies. Lastly, neither the severity nor frequency of the traumatic events in PTSD patients were considered as a possible covariate in this study. Further studies considering these effects should be executed in future. This is one of the first studies to examine the interaction effect of BDNF rs6265 and childhood trauma on both PTSD symptoms and cortical thickness in Korean population. We found that people with both the Val/Val genotype of rs6265 and high CTQ score showed severe PTSD symptoms. Furthermore, this interaction affected the thickness of both the left fusiform and transverse temporal gyri, which significantly correlated with psychological symptoms, although such relationship differed between genotypes. Finally, these results indicate the possibility that the Val/Val genotype may increase the risk of such effect in our Korean population.

Participants.
A total of 255 Korean volunteers were initially included in the present study. However, 12 participants were excluded due to missing data from psychological measurements; nine additional participants were omitted from the analyses due to missing MRI and rs6265 data; and 18 participants who showed high denial scores in the childhood trauma questionnaire were also excluded, leaving a final sample of 216 volunteers. While the healthy participants were 133 (61.57%) and recruited through community advertisements, the PTSD patients were 83 (38.43%) and were recruited through notices on the bulletin board in the hospital instead. The PTSD patients were diagnosed based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) by a psychiatrist, and healthy participants were also assessed using the DSM-5 by a psychiatrist. The PTSD patients were mainly exposed to the following traumatic events: 57 (68.67%) to severe motor vehicle accidents, 7 (8.43%) to defect from North Korea, other 7 (8.43%) to physical or sexual violence, further 7 (8.43%) to relationship issues, 4 (4.82%) to death of a family member, and 1 (1.20%) to fire. The total participants pool was comprised of 72 (33.33%) men and 144 (66.67%) women, with a mean age of 45. www.nature.com/scientificreports www.nature.com/scientificreports/ abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect, and another scale for detecting minimization and denial 79 . The CTQ is comprised of 28 items and it is assessed with a 5-point Likert scale, ranging from 1 ("never true") to 5 ("very often true").
To examine the severity of PTSD symptoms, the posttraumatic stress disorder checklist for DSM-5 (PCL-5) was administered. The PCL is a self-report rating scale for assessing symptoms of PTSD 80 and was well-validated in Korea 81 . It consists of 20 items and is rated using a 5-point Likert scale, ranging from 1 ("not at all") to 5 ("extremely"). In addition to PTSD symptoms, The Korean validated version of the posttraumatic growth inventory (PTGI) was conducted to measure the degree of reported positive changes experienced after traumatic events 82 . It comprise four subscales, namely changes of self-perception, increase of interpersonal depth, finding new possibilities, and increase of spiritual interest 83 . It consists of 21 items and is rated using a 6-point Likert scale, ranging from 0 ("I did not experience this change as a result of my crisis") to 5 ("I experienced this change to a very great degree as a result of my crisis").
To control the exposure to traumatic events other than childhood trauma, the Korean validated version of Life Events Checklist (LEC) was used to assess the experience of potentially traumatic events 84 . The LEC comprised of 17 items of PTEs and the responses include experiencing, witnessing, and learning about it. This study analyzed responses of experience since other responses could be confusing to some respondents.
Other psychological symptoms that could possibly comorbid with PTSD were also assessed. The Korean validated version of the hospital anxiety and depression scale (HADS) was used to evaluate anxiety and depression symptoms. It consists of 7 items for describing anxiety and 7 items for determining depression 85 , and it is assessed with a 4-point Likert scale, ranging from 0 (no problems) to 3 (maximum distress). Moreover, the Korean version of the insomnia severity index (ISI) was used to evaluate the difficulty in falling and staying asleep, problems in waking up too early, the satisfaction with current sleep patterns, the interference with daily functions, the perception of impairment attributed to sleep problems, and the distress caused by the sleep problem 86 . It comprises seven items assessed with a 5-point Likert scale ranging from 0 to 4, with a higher score indicating greater insomnia severity. Additionally, the Alcohol Use Disorders Identification Test (AUDIT) was used to assess alcohol consumption, drinking behaviors, and alcohol-related problems before and after experiencing traumatic events. The AUDIT is a 10-item screening tool developed by the World Health Organization, and well-validated in Korea 87 . The AUDIT is assessed with a 5-point Likert scale ranging from 0 ("never") to 4 ("4 or more times a week").
Questionnaires that measures cognitive styles which could affect the onset of PTSD were assessed as well. The Korean validated version of the ruminative response scale (RRS) was conducted to measure ruminative responses. It is comprised of three subscales, specifically evaluating self-reproach, reflection, and depressive rumination 88 . It consists of 19 items and is rated using a 4-point Likert scale, ranging from 1 ("almost never") to 4 ("almost always"). Furthermore, the Korean validated version of the cognitive emotion regulation questionnaire (CERQ) was applied to measure the specific cognitive emotion regulation strategies used by participants in response to stressful life events. It consists of nine subscales, evaluating putting into perspective, refocus of planning, positive refocusing, self-blame, blaming other, focus on thought rumination, acceptance, catastrophizing, and positive reappraisal 89 . It consists of 36 items and is assessed with a 5-point Likert scale, ranging from 1 ("almost never") to 5 ("almost always"). BDNF genotyping. All participants had their blood sampled to extract their DNA using a NanoDrop ® ND-1000UV-Vis Spectrophotometer. Thereafter, their genomic DNA was diluted to a 5 ng/µ concentration on 96 well PCR plates. TaqMan SNP Genotyping Assays were obtained from Applied Biosystems. Following, the probes were labeled with either the FAM or the VIC dye at the 5′ end and with a minor-groove binder and a non-fluorescent quencher at the 3′ end. PCR was performed in a 5 μl of a mixture containing 2 μl of a DNA sample, 0.125 μl of each TaqMan ™ SNP Genotyping Assay (Thermo Fisher Scientific, USA), 2.5 μl of the TaqMan ™ Genotyping Master Mix (Thermo Fisher Scientific, USA), and 0.375 μl of distilled water. Amplification and detection were completed with a detection system (QuantStudio 12 K Flex Real-Time PCR System, Thermo Fisher Scientific, USA), using the following profile: 50 °C for 2 min, 95 °C for 10 min followed by 60 cycles of 95 °C for 15 sec, and 60 °C for 1 min. Successively to PCR amplification, allelic discrimination, an endpoint plate read, is performed at the same machines (QuantStudio 12 K Flex Real-Time PCR System). In fact, the QuantStudio 12 K Flex Software calculates the fluorescence measured during the plate read and plots Rn values based on the signals coming from each well. Subsequently, automatic or manual allele calls were performed on the analyzed plates.
Three positive and one negative control samples were present on each plate. We confirmed the clustering image with positive controls, which consisted of. Intra genomic DNA (gDNA) samples of known genotypes are used for positive control.
Sixty-seven participants had the Val/Val (CC) genotype, whereas 149 participants reported the had Met allele (CT + TT) in the rs6265.
MRI acquisition and processing. A 1.5 T scanner (Magneton Avanto, Siemens, Erlangen, Germany) with restraining foam pads to reduce head motions was used for MRI acquisition. Parameters for MRI images were same as our previous study: a 227 × 384 acquisition matrix, a 210 × 250 field-of-view, 0.9 × 0.7 × 1.2 voxel size, a total of 87,168 voxels, a TE of 3.42 ms, a TR of 1,900 ms, 1.2 mm slice thickness, and a flip angle of 15°9 0 .
After visual inspection for artifacts, the surface-based morphometry (SBM) analysis was executed using the Computational Anatomy Toolbox (CAT12, http://www.neuro.uni-jena.de/cat/) toolbox in SPM12 (Wellcome Department of Cognitive Neurology, London, UK). The images were regularized with the International Consortium for Brain Mapping template for East Asian brains and normalized using the Diffeomorphic Anatomical Registration using Exponentiated Lie algebra (DARTEL) algorithm 91 . The images were segmented into gray matter, white matter, and cerebrospinal fluid 92 using Jacobian-transformed tissue probability maps. www.nature.com/scientificreports www.nature.com/scientificreports/ The cortical thickness was estimated with the projection-based thickness method 93 and extracted using the Desikan-Killany Atlas (DK40), which contains 35 cortical areas in each hemisphere 94 . The cortical thickness value at each vertex was calculated as the closest distance from the gray/white boundary to the gray/CSF boundary 95 .
Statistical Analysis. Normality was tested using the skewness and kurtosis measures. Skewness over 2.0 and kurtosis over 7.0 are considered to present moderately non-normal distribution 96 . All variables in our results were within the range of normal distribution. After checking for normality, a regression analysis using the SPSS Macro PROCESS for SPSS 2.16.3 97 was performed to examine the interaction between childhood trauma and rs6265 on the PCL score while controlling for the effect of gender, age, and years of education as covariates. Subsequently, the CTQ score was partitioned into two, at the point in which the moderation effect was first observed. A two-way multivariate analysis of covariance (MANCOVA) was used to find the interaction effect of childhood trauma and rs6265 on cortical thickness, whereas a correlation analysis was conducted to find the relation between cortical thickness and psychological symptoms. All statistical analyses were performed with the SPSS 25 (SPSS, Inc., Chicago, IL, USA). The order and the process of statistical analyses are presented in Table 4. All measurements and all experiment protocols were approved by the Inje University Ilsan Paik Hospital Institutional Review Board (IRB no. 2015-07-025) and were executed in accordance with guidelines and regulations of the board.

Data Availability
The data that support the findings of this study are available from the corresponding author, S.-H.L., upon reasonable request.

Process
Analyses Reasons for Analyses 1 Normality Testing to examine the interaction normal distribution for further analyses 2 Moderated Regression Analysis to examine the interaction effect of childhood trauma and rs6265 on PTSD symptoms and discover the score where the interaction happens 3 Multivariate Analysis of Covariance to examine the interaction effect of childhood trauma and rs6265 on cortical thickness based on the score of the interaction 4 Correlation Analysis to examine the relationship between psychological symptoms and cortical thickness of regions affected by the interaction of childhood trauma and rs6265 Table 4. the process of analyses and reasons for statistical methods.